Audio-Video Based Emotion Recognition Using Minimum Cost Flow Algorithm

Nguyen, Xuan-Bac; Lee, Guee-Sang; Kim, Soo-Hyung; Yang, Hyung-Jeong

doi:10.1109/iccvw.2019.00464

Cited by 3 publications

(3 citation statements)

References 7 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Our methods, including CNN, LSTM, and the combined CNN-LSTM model, outperformed all other methods, exhibiting accuracy rates of 90.56%, 90.90%, and 94.94%, respectively. [18] 84.4% Xu et al [21] 80.5% Nguyen-Xuan [22] 90.4% Kothawade et al [4] 65.57% Guo et al [23] 60% Our CNN 90.56% Our LSTM 90.90% Our CNN-LSTM 94.94%…”

Section: Cnn-lstm Resultsmentioning

confidence: 99%

“…They also emphasized the investigation of temporal feature ordering by employing a two-branch network that combines both a CNN and RNN. Nguyen et al [22] presented a novel sequential model that incorporates a combination of CNN and LSTM along with attentional mechanisms [14]. This model has been applied in the field of image analysis, where researchers have employed multiple pretrained models in ImageNet to identify doodles.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

A Combined Deep Cnn-Lstm Network For Sketch Recognition

EL Mouna,

Silkan,

C´edric Koumetio Tekouabou

et al. 2024

IJCDS

View full text Add to dashboard Cite

Automating freehand sketches is a complex process due to their diverse and abstract characteristics. Recently, there has been significant interest among researchers in machine learning algorithms, owing to their emergence. Nevertheless, many utilized models are either inadequate or overly complex, featuring processes that lack clarity and consistency, which hinders their ability to accurately depict real-world scenarios. In this study, we introduce an approach that applies deep learning methods involving a combination of Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) to enhance sketch recognition performance. In the initial phase of our approach, a CNN was employed to extract features that were subsequently forwarded to an LSTM network for classification. We evaluated the efficacy of our method by utilizing the QuickDraw dataset offered by Google, and the results demonstrated that our approach outperformed both CNN and LSTM, as well as other state-of-the-art methods. Our method attained an accuracy of 95%, with precision and recall reaching 95%, while also achieving an F1 score of 94%.

show abstract

Section: Cnn-lstm Resultsmentioning

confidence: 99%

Section: Related Workmentioning

confidence: 99%

A Combined Deep Cnn-Lstm Network For Sketch Recognition

EL Mouna,

Silkan,

C´edric Koumetio Tekouabou

et al. 2024

IJCDS

View full text Add to dashboard Cite

show abstract

“…These include traditional techniques [49,50], graph-based methodologies [42][43][44][45][46]51], and transformer-based approaches [35]. While transformer architectures have demonstrated significant success in various computer vision tasks [35,48,[52][53][54][55][56][57][58][59][60][61][62][63][64][65], their potential in quantum computing remains promising. Adapting the typical transformer architecture for quantum systems, as proposed by [66], offers added convenience.…”

Section: Experiments Setupmentioning

confidence: 99%

Quantum Visual Feature Encoding Revisited

Nguyen,

Churchill

et al. 2024

Preprint

View full text Add to dashboard Cite

Although quantum machine learning has been introduced for a while, its applications in computer vision are still limited. This paper, therefore, revisits the quantum visual encoding strategies, the initial step in quantum machine learning. Investigating the root cause, we uncover that the existing quantum encoding design fails to ensure information preservation of the visual features after the encoding process, thus complicating the learning process of the quantum machine learning models. In particular, the problem, termed ”Quantum Information Gap” (QIG), leads to a gap of information between classical and corresponding quantum features. We provide theoretical proof and practical examples with visualization for that found and underscore the significance of QIG, as it directly impacts the performance of quantum machine learning algorithms. To tackle this challenge, we introduce a simple but efficient new loss function named Quantum Information Preserving (QIP) to minimize this gap, resulting in enhanced performance of quantum machine learning algorithms. Extensive experiments validate the effectiveness of our approach, showcasing superior performance compared to current methodologies and consistently achieving state-of-the-art results in quantum modeling.

show abstract

Quantum visual feature encoding revisited

Nguyen,

Churchill

et al. 2024

Quantum Mach. Intell.

View full text Add to dashboard Cite

Although quantum machine learning has been introduced for a while, its applications in computer vision are still limited. This paper, therefore, revisits the quantum visual encoding strategies, the initial step in quantum machine learning. Investigating the root cause, we uncover that the existing quantum encoding design fails to ensure information preservation of the visual features after the encoding process, thus complicating the learning process of the quantum machine learning models. In particular, the problem, termed the “Quantum Information Gap” (QIG), leads to an information gap between classical and corresponding quantum features. We provide theoretical proof and practical examples with visualization for that found and underscore the significance of QIG, as it directly impacts the performance of quantum machine learning algorithms. To tackle this challenge, we introduce a simple but efficient new loss function named Quantum Information Preserving (QIP) to minimize this gap, resulting in enhanced performance of quantum machine learning algorithms. Extensive experiments validate the effectiveness of our approach, showcasing superior performance compared to current methodologies and consistently achieving state-of-the-art results in quantum modeling.

show abstract

Audio-Video Based Emotion Recognition Using Minimum Cost Flow Algorithm

Cited by 3 publications

References 7 publications

A Combined Deep Cnn-Lstm Network For Sketch Recognition

A Combined Deep Cnn-Lstm Network For Sketch Recognition

Quantum Visual Feature Encoding Revisited

Quantum visual feature encoding revisited

Contact Info

Product

Resources

About